NBER WORKING PAPER SERIES HOW DO WE CHOOSE OUR IDENTITY? A REVEALED PREFERENCE APPROACH USING FOOD CONSUMPTION David Atkin Eve Colson-Sihra Moses Shayo Working Paper 25693 http://www.nber.org/papers/w25693 NATIONAL BUREAU OF ECONOMIC RESEARCH 1050 Massachusetts Avenue Cambridge, MA 02138 March 2019 We thank Ben Deaner, Omer Karaduman, and Sumit Shinde for excellent research assistance. David Atkin thanks the Stanford Economics Department and SIEPR for their hospitality while writing this paper. Moses Shayo thanks the I-Core Program at the Israel Science Foundation (grant no. 1821/12) and the Falk Institute for financial support. We thank Roland Benabou, Jon Eguia, Armin Falk, David Genesove, Marco Gonzalez-Navarro, Matthew Gentzkow, Ori Heffetz, Supreet Kaur, David Laitin, Edward Lazear, ShabanaMitra, SharunMukand, Salvatore Nunnari, Ran Shorrer, Katia Zhuravskaya, the editors and four anonymous referees, and many seminar participants for valuable comments. The views expressed herein are those of the authors and do not necessarily reflect the views of the National Bureau of Economic Research. NBER working papers are circulated for discussion and comment purposes. They have not been peer- reviewed or been subject to the review by the NBER Board of Directors that accompanies official NBER publications. © 2019 by David Atkin, Eve Colson-Sihra, and Moses Shayo. All rights reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission provided that full credit, including © notice, is given to the source. How Do We Choose Our Identity? A Revealed Preference Approach Using Food Consumption David Atkin, Eve Colson-Sihra, and Moses Shayo NBER Working Paper No. 25693 March 2019, Revised June 2020 JEL No. D12,D74,D91,O1,Z1 ABSTRACT Are identities fungible? How do people come to identify with specific groups? This paper proposes a revealed preference approach, using food consumption to uncover ethnic and religious identity choices in India. We first show that consumption of identity goods (e.g. beef and pork) responds to forces suggested by social-identity research: group status and group salience, with the latter proxied by inter-group conflict. Moreover, identity choices respond to the cost of following the group’s prescribed behaviors. We propose and estimate a modified demand system to quantify the identity changes that followed India’s 1991 economic reforms. Notably, our estimated identity changes correlate with changes in vote shares for ethnic and religious parties. While social-identity research has focused on status and salience, our results suggest that economic costs also play an important role. David Atkin MIT Department of Economics The Morris and Sophie Chang Building, E52-550 50 Memorial Drive Cambridge, MA 02142 and NBER [email protected] Eve Colson-Sihra Department of Economics The Hebrew University of Jerusalem Jerusalem 91905, Israel [email protected] Moses Shayo Department of Economics The Hebrew University of Jerusalem Jerusalem 91905, Israel [email protected] A data appendix is available at http://www.nber.org/data-appendix/w25693 1 Introduction This paper asks whether people’s social identities are fungible, and if they are, how do people come to identify with specific groups? In their seminal paper, Akerlof and Kranton (2000) de- tail how identities affect a host of outcomes of interest to economists, including human capital acquisition, labor market participation and poverty. The important role identity plays is also salient in the current political environment. Both commentators and scholars have linked the ef- fects of globalization, immigration and rising inequality to a shift away from a now lower-status working class identity towards a nationalist one across both Europe and the US (Shayo 2009). Such identity shifts have also been linked to changes in trade policies and opposition to global- ization (Grossman and Helpman Forthcoming), with the political repercussions seen in Brexit, the resurgence of the far right in Europe, and Donald Trump’s election victory. Similar forces ap- pear to be playing out in India with the ascent of Narendra Modi and the Bharatiya Janata Party (BJP) on the back of a surge in Hindu nationalism. While it is straightforward to motivate why it is interesting to study identity choices and how they affect economic behavior, measuring identity choices is challenging. Identity is usually conceptualized in terms of preferences, i.e., it is part of the “self” whose self-interest we seek to maximize. Thus, to identify with different groups means to care about different things (for reviews see Haslam and Ellemers 2005; Charness and Chen 2020; Shayo 2020). However, to measure identity outside the lab, much of the existing literature relies on survey questions or on ethnographic and historical case studies. But while case studies can provide important in- sights, they are often very specific and evaluating causal arguments is hard. Surveys offer the possibility of broad representative samples, but it is at present less clear to what extent differ- ent self-reported identity measures reflect day-to-day economic behavior.1 Finally, experiments provide rich revealed-preference data but they are necessarily limited in scope to a particular time, place, and population (often students at American universities). A more recent strand of the literature uses naturally occurring data to understand how social identities can affect the be- havior of judges (Shayo and Zussman 2011), team production (Hjort 2014), female labor supply (Bertrand et al. 2015), grading decisions (Feld et al. 2016; Lavy et al. 2018), and conflict (Depetris- Chauvin and Durante 2019). We build on these recent advances. Our approach starts from the observation that consumption choices are both widely docu- 1With the increasing interest in identity and polarization, survey questions on identity are likely to become more common. Our approach suggests a method for validating such questions in the spirit of Falk et al. (2018)—by testing whether they capture non-hypothetical, costly decisions. For studies correlating survey and behavioral measures of identity see Ellemers et al. (1999); Klor and Shayo (2010); Bankert et al. (2017). 1 mented and are affected by the norms and taboos of groups people identify with. Since different groups have different norms, consumption choices have the potential to reveal the consumer’s identity. For example, conditional on prices and income, a Muslim who consumes pork iden- tifies less with his religion than a Muslim who abstains from pork. We explore whether this in- sight can help better understand identity choices. By drawing on standard and readily-available consumption data as well as well-established tools for analyzing them, we can investigate mul- tiple determinants of identity among a large and representative population over a long period of time. Turning the question around, we also ask whether ideas from social identity research can improve economists’ understanding of consumer behavior. Our setting is food consumption in India where, given the high levels of malnutrition, dietary choices driven by identity also have significant health implications. Two features make the In- dian setting particularly suitable for implementing our approach. First, India is characterized by deep ethnic and linguistic divisions (Basu et al. 2016; Fearon 2003; Reich et al. 2009). At the same time, it is also religiously diverse, with members of the same ethnicity often distributed across different religions and castes. This provides well-defined sets of (potential) identities that indi- viduals can choose from. Second, food consumption in India is associated with strong norms and taboos. In the words of noted anthropologist Arjun Appadurai (1988): Food in India is closely tied to the moral and social status of individuals and groups. Food taboos and prescriptions divide men from women, gods from humans, upper from lower castes, one sect from another. The basic idea then is simple. A Hindu from Gujarat is born a member of multiple groups and hence has several possible identities. While they cannot (easily) choose to be Muslim or Tamil, they can choose whether to identify as Gujarati (one of India’s many ethno-linguistic groups) or Hindu (one of India’s major religions). Given the different norms and taboos across religious and ethnic groups, the food consumption bundle uncovers this identity choice. Our analysis exam- ines how these revealed identities respond to key forces that economics and social psychology conjecture drive identity choices. We base our analysis around a conceptual framework that we outline in Section 2. Follow- ing the identity economics literature (e.g. Akerlof and Kranton 2000), an individual that iden- tifies with a group suffers a utility loss if her consumption bundle is far from the prescribed or prototypical bundle of that group. In other words, identification with a group means, among other things, seeking to consume the goods that are acceptable to that group. However, extend- ing Akerlof and Kranton (2000), we allow individuals to choose their identity among the multi- ple groups they are members of, depending on which provides the highest utility. We identify 2 three major factors that affect this choice. From cognitive psychology and categorization theory (Nosofsky 1992; Turner et al. 1987) as well as experimental economics (Benjamin et al. 2010), we posit that if the salience of membership of a group rises, the utility from identifying with